feat(llm): add native provider packages

[nexxeln]

Summary

• define the uniform model(id, settings) provider package contract used by the agreed config shape
• implement that contract directly in the existing OpenAI, Anthropic, OpenAI-compatible, and Amazon Bedrock provider modules
• map package headers, body, and limits onto the existing executable route configuration
• expose explicit OpenAI Responses and Chat package paths while keeping the root OpenAI path as the stable Responses default

This only establishes the LLM-side boundary for a mechanical core call:

module.model(effective.id, {
  ...effective.settings,
  headers: effective.headers,
  body: effective.body,
  limits: model.limit,
})

Core config parsing, overlay merging, package loading, and credential injection will follow separately. No new provider or protocol implementation is introduced here.

Testing

• bun typecheck from packages/llm
• bun test --timeout 30000 --only-failures from packages/llm
• repository pre-push typecheck across 25 packages

Stack

1. feat(llm): add native provider packages #33689 👈 current
2. feat(core): migrate native provider config #33697